GT Axis Daily, July 30, 2026

A statement called "Pacing the Frontier" gathered 1,268 verified employees from OpenAI, Anthropic, Google DeepMind, and Meta. They are not asking for a pause today. They are asking Washington to help build the international tools that would make a coordinated slowdown possible later, specifically if AI starts improving itself faster than people can supervise it. Within hours, OpenAI and Anthropic endorsed it at the company level, with Anthropic CEO Dario Amodei and OpenAI chief scientist Jakub Pachocki among the signatories. The split is the interesting part: Meta's own chief AI scientist signed as an individual while Meta joined NVIDIA, Microsoft, and Palantir on a separate letter asking regulators not to restrict open model releases. The industry is no longer speaking with one voice on its own rules.

The program starts with 10,000 researchers this summer, including at the Institute for Advanced Study and École normale supérieure, and scales toward 100,000 by 2027. Each researcher gets frontier model access plus four colleague invitations, in a workspace where their data is not used for training by default. It is part of a commitment OpenAI puts at more than $250 million through 2027. Read it as two things at once: real support for science, and the cheapest possible way to make sure the next generation of researchers is fluent in your product before anyone else's.

Quarterly revenue hit $90 billion, up 18%, with Azure growing 43% in the quarter and passing $100 billion for the full fiscal year. Its Copilot assistant reached over 30 million paid seats, which matters because paid seats are recurring revenue rather than a free trial number. Microsoft also added 88 data centers during the year and guided to roughly $175 billion in capital spending next fiscal year, a figure that would have terrified investors a week ago. It did not today, because the revenue line moved with it. Management still describes itself as capacity constrained, meaning it could sell more if it had somewhere to run it.

Revenue came in at $60.8 billion, up 28% and ahead of expectations, but capital expenditures hit $31.1 billion in a single quarter, nearly double a year ago. The number that actually broke the stock was free cash flow, the cash left after running and building the business, which fell to $784 million from $8.55 billion a year earlier. Meta then narrowed full-year capital spending guidance to $130 to $145 billion by raising the floor, not lowering the ceiling. Investors heard a company committing to spend more while generating almost nothing spare.

Apple's call is Tim Cook's last as chief executive, which means the outlook language carries more weight than usual. Analysts expect roughly $109 billion in revenue and about $1.89 per share, with the real attention on iPhone demand, services growth, and how much Apple plans to spend on AI given it rents most of its model capability rather than building it. Amazon's report lands the same afternoon, where AWS growth and data center spending are the two lines that matter.

Kilo Code is an open-source AI coding assistant that works across VS Code, JetBrains, the browser, and the command line, and it is model-agnostic, meaning it routes work to whichever AI model you choose rather than locking you into one vendor. It carries more than three million users and pushes close to ten trillion tokens a month across over 500 different models. Anaconda, which already supplies the Python tooling inside most large enterprises, is stitching it into a governance layer so companies can see and control which model their developers are actually sending code to. The unglamorous read: the product being sold here is not the assistant, it is the audit trail.

The $1.8 billion acquisition of SkyWater Technology completes July 31. SkyWater is a DMEA Trusted Foundry, meaning a US chip plant cleared by the Defense Department for sensitive work, and owning it makes IonQ the first quantum company controlling its own design, wafer fabrication, packaging, and system assembly under one roof. Every other quantum name still depends on outside foundries and foreign supply chains for its hardware. IonQ just stopped. SkyWater keeps serving its existing customers as a subsidiary, so the deal also brings real revenue rather than pure research burn.

SoftBank exercised a put option written into the 2021 deal, a contract clause that let it force Hyundai to buy its remaining stake if the robotics company had not gone public by this June. Hyundai paid roughly $325 million for the last 10%, which values Boston Dynamics at about $3.3 billion, the same number as five years ago. Sit with that: the most famous robotics company on earth produced zero value appreciation over five years while the rest of AI went vertical. Hyundai says it will fold the company into an integrated robotics operation across its factories and supply chain, which is the only path that makes the price make sense.

The FOMC voted 9 to 3 to keep the federal funds rate at 3.50% to 3.75%, paired with firm language on persistent inflation. Bond markets took it badly, with the 30-year Treasury yield reaching levels not seen since 2007 as traders gave up on near-term cuts. Here is why a leader should care: AI data centers are built with borrowed money against long leases, and the cost of that borrowing is set by exactly these yields. Capital was never the shortage in this buildout. Cheap capital was.

The company reported its fiscal fourth quarter with roughly $36 billion in total contracted revenue across five campuses and about 1,410 megawatts of committed capacity. The headline additions were three fifteen-year take-or-pay leases totaling 810 megawatts with a single high investment-grade hyperscaler. Take-or-pay means the tenant owes the money whether or not it uses the capacity, which turns a construction project into something closer to a bond. Full-year revenue grew 167%, and the company still is not profitable, which is the honest caveat.